High Dynamic Range Imaging

Instructor: Jiang Duan, Southwestern University of Finance and Economics, China

Introduction

The dynamic Range of a scene, image or imaging device is defined as the ratio between the highest to the lowest luminance or signal level. The real world thus has a dynamic range of approximate fourteen orders of magnitude. The current performance of digital image capture and display devices can match or even exceed those of silver halide film in many aspects like accuracy and resolution. However, these devices still suffer from a limited dynamic range, typically spanning only two to three orders of magnitude. We call these devices Low Dynamic Range (LDR) reproduction devices.

Recently there has been increasing interests in High Dynamic Range Imaging (HDRI) to better capture and reproduce real scene. In HDRI, the image files use 32-bit float point value per channel to faithfully record the full dynamic range and the actual color gamut of the real scene. In comparison with conventional images which only use 8-bit integer value per channel, we can much more precisely store, edit, manipulate and display images in HDRI workflow and greatly improve users’ imaging experience.

Because of the advantage of HDRI, HDRI technologies are widely used in many fields like digital photography, medical imaging, digital cinema and computer games, etc and will be possibly used in many new areas like archaeology, astronomy, etc. In this tutorial, we will cover related topics of high dynamic range imaging technologies which include the creation of radiance map, the encoding of radiance map, the tone mapping technologies, high dynamic range (HDR) monitors and HDRI applications. The tutorial will give the audience a full picture of High Dynamic Range Imaging and will last half a day.

Content Sketch

Background

We will discuss the limitation of current Low Dynamic Range imaging devices and thus will explain why we need High Dynamic Ranging Imaging technologies. In addition, in this part, we will briefly introduce relevant knowledge about adaptation mechanism in Human Visual System and explain why our eye can simultaneously see details in all parts in high dynamic range scene and why conventional imaging devices cannot.

HDRI acquisition

Currently, HDRI acquisition is achieved by merging a set of photos with different exposures. To get this done properly, we will discuss the following technologies.

  1. Alignment methods for different exposed photos It is unavoidable that translational, rotational movement and even distortions among the bracketed images may occur during the capture process. Relevant techniques of aligning multi-exposed photos will be introduced.
  2. Ghost removal technologies Sometimes, objects’ movement exists in the scene at the time of capture and will appear as ghost artifacts in the merged images. Relevant techniques of removing ghost artifacts will be introduced.
  3. Noise removal Noises are usually presented for low lighting photos and thus presented in the merged radiance map. We will discuss the technique of noise reduction techniques in HDRI workflow
  4. The generation of radiance map Different exposed photos need to be merged to 32-bit radiance map to cover the full dynamic range and color gamut of the real scene. Relevant techniques of recovering radiance map will be introduced.

HDRI coding and file formats

32-bit radiance map needs to be encoded for efficient storage and transmission with enough accuracy. The representatives of these encoding are IEEE TIFF Format that uses lossless encoding method, RGBE Format that uses the lossy encoding method and is also the most widely used format in the industry and SGI LogLuv TIFF format that is based on the human perception. We will discuss these techniques in this part.

HDR image display/reproduction

32-bit radiance map cannot be displayed directly on conventional visualization devices and thus specific methods or devices are needed to present 32-bit radiance map. We will discuss the techniques of hardware and equipment used to display 32-bit radiance map directly, which includes the method of Seetzen et al and an introduction to Brightside HDR monitors. We will discuss tone mapping or tone reproduction techniques used to render 32-bit radiance map to display on conventional visualization devices in such a way that the visual impressions and feature details of the original real scenes are still faithfully reproduced. In the literature, tone reproduction techniques for compressing dynamic range are usually described in two broad categories: global and local tone mapping operators. We will explain the difference between both categories and will give a comprehensive introduction of important methods from both categories.

HDRI application

The emphasis on this part is one of the important applications of HDRI ---- HDR photography. We will introduce the workflow and software tools needed for HDR photography, we will also teach audiences how to make beautiful HDR photos. We will also introduce other HDRI applications and its future possible application in the field of medical imaging, digital cinema and computer game and astronomy, etc.